Yes, I worked that out.
In a 2716 as used in that project, the address range is 000-7FF, so address 800 does not exist.
I have checked through exactly half of the generated code so far against a large A3 sized printout of the original dump in the article. I've found several more recurring errors including the occasional 'S' where '5' should be - that was caught by the ASCII hex to raw binary conversion process because 'S' is not a valid hex character - and also a leading 'D' in an ASCII byte was sometimes mistaken for '0' at the original OCR stage. Unfortunately both 'D' and '0' are valid hex digits so I'm having to root those ones out manually.
I'm assuming I can take my time to do the rest, as I don't imagine you'll have the hardware built in the next five minutes.
If, by the time I do this, you still don't have the means to read a code file into one of your programmers and then program it into an eprom, I could of course send it to you -
in an eprom-. You could then read it out into your PP28 and step through it checking my (hopefully correct) version against your own printout.
If you were happy that it was correct you could just use the eprom as-is, but if you found that further errors needed correcting you could change the offending data in the programmer's RAM and then program the amended code into another blank 2716.
As far as the ability to load and use code files is concerned, the Dataman is still the better bet, and I would urge you to try installing the PC support software for that and have a go with it - Programmers driven from a PC are always far easier to use than standalone programmers.